[Ml-stat-talks] Seminar, Han Liu, Thu 2/16, 4:30 PM

Ahmet Emre Barut abarut at Princeton.EDU
Thu Feb 16 10:34:04 EST 2012

Dear all,

Han Liu from the Department of Biostatistics and Computer Science at Johns Hopkins University<http://www.cs.jhu.edu/~hanliu/> will be giving a talk today (February 16) at 4:30 PM in Sherrerd 101, as part of the ORFE Colloquium. He works on a wide array of statistical problems, and today he will present his work on graph estimation using nonparametric methods.


P.S. To the best of my knowledge, he is also the first person ever to get a PhD in Machine Learning AND Statistics.

TITLE:  Nonparametric Graph Estimation

ABSTRACT: The graphical model has proven to be a useful abstraction in statistics and machine learning. The starting point is the graph of a distribution. While often the graph is assumed given, we are interested in estimating the graph from data. In this talk we present new nonparametric and semiparametric methods for graph estimation. One approach is a nonparametric extension of the Gaussian graphical models that allows arbitrary graphs. Another approach is to restrict the family of allowed graphs to spanning forests, enabling the use of fully nonparametric density estimation in high dimensions.  These two approaches can both be viewed as special cases of the more general log-density ANOVA models and reflect an interesting tradeoff of model flexibility with structural complexity. In terms of function estimation, these methods achieve the minimax optimal rates of convergence. In terms of graph estimation, these methods even achieve the optimal parametric rates of convergence. Therefore, the extra flexibility gained by nonparametric modeling comes at very low cost. In terms of computing, we provide currently the most scalable software package that is several times faster than the state-of-the-art softwares implementing the standard parametric methods. The performance of these methods is illustrated and compared on several real and simulated examples. Joint work with Fang Han, John Lafferty, Larry Wasserman, and Tuo Zhao.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/ml-stat-talks/attachments/20120216/2f763c38/attachment.htm>

More information about the Ml-stat-talks mailing list